Display calibration. Part V.

For a better understanding of the terms and concepts I will mention in this article, allow me to recommend reading the third chapter of my colleague Javier Guerra's calibration special, published in issue no. 67 of July 2009, which covers in detail the theory needed to easily understand what we will see below.

In the article on basic settings, I deliberately avoided discussing the adjustment of the "color" control, found on virtually every display. I have not mentioned it until now for the simple reason that I consider it relevant only in cases where there is no CMS with controls for adjusting the primary and secondary colors and, even then, only as a "coarse" complement to them. On the other hand, even if the only means of color adjustment available is that control, filters should not be used, because of their lack of precision and because, as we will see below, we can do the image more harm than good.

Most manufacturers implement the color control so that it simultaneously modifies the saturation and luminance of all colors, acting as a "chroma" control. If all colors behaved similarly on our display, it would save us time when calibrating, eliminating the need to adjust each one independently. But the reality is that I have yet to find a display where this happens. In fact, it is extremely common to find that while some colors have excess saturation or luminance, others show greater or lesser shortfalls. If we have no CMS of any kind, our only way of correcting deviations will be to use this control and, as on previous occasions, we will have to look for "compromise" solutions, that is, those that minimize errors, while knowing that the result will not be optimal. By way of example, let's imagine that our display had only two colors, red and green, the first with excess saturation and luminance, and the second behaving in exactly the opposite way. In this situation, we could optimize red at the cost of increasing the green error even further, or look for a point at which both are "less bad" but neither is right. This seemingly simple decision (we would go for the second option) becomes more complicated depending on the differences in error, and when there are not two but six colors to consider. To know the exact amount by which the default value must be changed, or even whether it is better to leave it alone, there is no alternative but to use calibration equipment.

Unlike HCFR, applications such as Calman or Chromapure let us see to what extent each component contributes to the overall deviation ΔE, which for convenience we will call dE, replacing the Greek letter delta with its initial. There is nothing better than an image to explain this section, in this case a screenshot from Spectracal's Calman software (Pioneer PDP-LX5090H):



If in this screenshot we look at the vertical section containing the boxes for the different deltas -at the top, the Total, and just below, those for Luminance (which from now on we will call L, and its deviation dEL), Color (S, dES) and Hue (H, dEH)-, we can see two good examples of how each component contributes to a different extent depending on how it influences our perception. Looking at two colors with a similar Total dE, blue and magenta, we can clearly see the importance of dEL: in the case of blue, thanks to a low figure for this error, it can "afford" relatively large dES and dEH errors and yet the final aggregate figure does not reach a value of 2; whereas with magenta we can see how the Total dE almost matches the dEL, however low its dES and dEL may be.

In the center, the CIE diagram stands out, in this case the 1931 xy model, which can be switched to the CIE UV model, considered more perceptually uniform, meaning that it shows as larger those errors that our eyes perceive more easily. If we switch to this model in the previous example, we get the following:



With this form of representation it is easier to see that red, magenta and blue are further from their targets than they appeared in the CIE 1931 model.

At the top right is the table containing the percentage deviations in luminance relative to what each color should have, based on the white measured with a pattern of the same intensity. In our example, magenta has a luminance 7% lower than it should.

Once we have understood how to use the screen found in the gamut adjustment section of our application, the next thing to consider is selecting the intensity of the patterns to be used for color adjustment. If we look for information on the Internet about this, we find conflicting opinions on whether to use 75% or 100% intensity patterns, with strong arguments for and against both options. Personally, I prefer to use the former (since they are potentially closer to the middle of the dynamic range), except on those occasions when the display does not behave well at that intensity, which is visually more critical than 100%, in which case we will use the latter.

One concept that must be very clear is that CMS adjustments always reduce the color gamut of our display; logically, it is not possible to increase it beyond its physical capabilities. Not even with the help of external processors can we go beyond the physical limits of the panel or projector. For this reason, whenever we have a complete and functional CMS, it is advisable to choose the "extended" color modes if the "normal" or "native" modes cover a smaller area than the standard we are calibrating to. Using the CIE UV chart as an example, we can see that blue lies outside the target area. In this case, it would not be possible to move blue to its ideal coordinate, since it is outside the display's limits. It would, however, be possible to move it along the sides of the triangle or even inward, at the cost of reducing our color range. Another side effect of gamut adjustment can be the appearance of banding and posterization due to rounding (if it is carried out with 8-bit precision or less) and the reduction in dynamic range resulting from lowering the RGB values of each color. The higher the quality of the equipment used (processing capability, control design, etc.), the less these types of problems will appear.

In our example (see the complete chart), the saturations are reasonably good: green is very slightly oversaturated, red somewhat exaggerated, and blue is not oversaturated but is shifted toward green. At the same time, we can see that green has excess luminance, while the opposite is true of red (blue is basically correct). If we only have the "color" control common to all displays, there is little we can do since, as I said earlier, because it affects saturation and luminance simultaneously, whatever we correct on one side we will spoil on the other. If we lower "color", we will reduce the saturation of red, which will look better on our CIE chart but will lose too much luminance. Green will fall inside the target CIE area and will improve as its excess light is reduced, while blue will acquire a luminance error that it hardly has now, and its Total dE will increase noticeably.

There are numerous implementations of CMS sets; each manufacturer understands adjustment in its own way (and there may also be patents) and implements a different number of controls. Some work better than others, and in some cases it is better not to use them at all, because they introduce side effects that degrade the image more than any possible improvement, or they don't do what they are supposed to do. We may find, as in the case of the JVC HD750, a 3D CMS in which the names of the controls match the parameter to be adjusted (L, S and H for each primary and secondary color) or, as in the case of Samsung and Lumagen video processors (both of which work excellently), the ability to alter the RGB combination for each of those same colors, in which the component of the same color (e.g. B in blue) acts as L, changing the complementary components by different amounts acts as H, and by the same amount as S. There are also displays with a 2D CMS, such as those used in LG or Panasonic panels (primary colors only), which only offer S and H adjustment (we can use the "color" control together with S to fine-tune). In any case, and regardless of the CMS we come across, its operation must always be verified using the "measure ->small adjustment ->measure" process.

Despite the effort made to condense and make "digestible" a subject that could fill much more space, what was initially going to be a two-part miniseries has ended up becoming a five-part saga, with which I hope I have helped encourage you to consider adjusting your own displays and getting the very best performance they are capable of delivering.